Jump to: navigation, search

PTP/planning/2.0

< PTP‎ | planning
Revision as of 14:54, 14 March 2008 by Tibbitts.us.ibm.com (Talk | contribs)

(diff) ← Older revision | Latest revision (diff) | Newer revision → (diff)

Current TODO List

This is the current TODO list for 2.0

Resource Manager Support

This will integrate the parallel model and parallel views with resource management systems. Users will be able to view and query the status of queues, and submit jobs through resource managers.

  • Submit and control jobs (viewing of jobs planned for previous release, 1.1)
  • Second implementation of a resource manager, possibly LoadLeveler?

Remote Services

Remote services support will provide the following functionality:

  • Allow projects to reside on remote systems
  • Ability to build projects remotely
  • Ability to launch and debug projects remotely

Parallel Debugger

  • Scalability improvements (10K processes)
  • Support for non-gdb backend debuggers (e.g. Absoft FX2)
  • Improved debugger functionality
  • New user interface features, including multi-variable viewer and array viewer
  • Visualization features

Additional Runtimes

  • Support for one additional parallel runtime system.

PLDT

  • New UI uses single toolbar menu
  • MPI Barrier Analysis detects potential deadlocks
  • New Project Wizards for MPI (more) and OpenMP (new)
  • New Project Wizards allow more settings e.g. build & link commands

Photran

  • CDT DOM
  • Error Checker (extended)
  • Refactorings (suggested by users)

Performance Analysis Framework

An implementation of an initial performance extension for ptp has been done for TAU, and is available in the PTP CVS repository, under org.eclipse.ptp/tools/tau.

Currently under development (July '07) is a generalized performance framework to assist tool developers in porting their tools to Eclipse, or easily integrating existing performance tools with Eclipse.

Integration points are envisioned to include:

  1. instrumentation, automatic and selective
  2. build, which may or may not be transparent to user
  3. launch with instrumentation
  4. management of profile/trace data
  5. visualization launch

Performance Tools Framework Design has more details

See also Performance Analysis Use Cases at [1]